### Next steps: - Cart testing - Jeremy, Paulo, Zuzana, Shelley on personalization signals data - [ ] UPD page testing - Discuss with Rohit, Vidhi on next steps re-using the experimental UX - [ ] Zuzana to update with test setup and research analysis on stockout nudge ##### TNT nudge experiment & Rohit Goel discussion - TNT confirmed that hard coded text can be shown based on historical data - Analyze 5 top products by purchases and views to show nudges on. - Decide which products to show nudges on. - Decide what types of nudges to show. - Popularity nudges (purchases, views) Rohit Goel - Ripa Patel’s team, Felisha’s team on AI-ML initiatives - Enable signals - PD pages takes most traffic and generates most revenue - We are not offering few eCom features that competitors are offering - Justin Walker, everybody liked it, funding conversations haven’t happened yet - Discussed with Suman (architect) - Justin is aware that TnT cannot test it ### Sizing Unified Cart Sizing - 4 sprints, 2 dev - Assuming new API integration, new API contract Insights svc Crawl 1. Index carts Elastic search index Geo-Partitioned Cart search AMER(only for MVP) Db - 1 sprint for 2 dev Indexing existing Carts for top 5 ordercodes. Active Carts or Carts that are ordered. - 2 sprints 2 dev 2. Search carts API Graph QL new endpoints to read and write - 3 sprints 2 dev Pre-compute the applicable nudges for carts 3. TxInsights API - client exposed API Input with Cart ID, item ID, path, signals --> Returns the nudge Live call to Search carts API to compute nudge 2 sprints 2 dev 1. Events: 3 sprints 2 dev On item added/deleted events Cart consumer listens and re-index Cart ordered event 2. Cart search DB cleanup job/utility - 1 sprint 2 dev Total: 14 sprints 2 dev. Some of these can happen in parallel and also number of devs will impact total no. of sprints If done in parallel: 8 sprints, 4 dev Walk - Index carts - Type sense vs Elastic search Db - 1 sprint for 2 dev - Configuration API for events configurations - 3 sprints 2 dev -Pre populate cart search indexes with batch job -Search API for carts data -Eventing in place for item added/deleted, cart ordered --> Search API for applicable carts --> Update applicable carts nudges -Nudge system with TxInsights API which accepts Cart ID, item ID, path, signals --> Returns cart ID and nudge details ### About Transact Insights PPT ##### Why not be like Apple? - Apple doesn't send emails on abandoned carts. What about our abandoned cart email reminders? - Apple is in Premium price bands only. What about our lower price band products? ##### Introduction - We see content as the biggest DSAT across Dell. During purchase, top 3 reasons customers are dissatisfied are - unable to find right product information, configure correctly and compare products. - We experimented recently to solve this by creating badges like top rated monitor, best UltraSharp, etc. and successfully proved that this tactic works from financial and CX improvement perspective. - Also, from external research on e-commerce purchase behavior out there, it shows that dynamic content that presents a relevant nudge in a positive manner helps boost user confidence and positively affects metrics such as conversion rate and WTP. - I see that just CSB alone can benefit by up to $120M incremental revenue if we address CR problem. Plus it’ll improve CSAT, that’ll be addition $ impact on top. - Here’s how we want to build an AI centric system to tackle this. - Preface everyone with this: - Explain that nudge doesn't mean FOMO tactics. Nudge as a concept exists to influence user behavior predictably towards a desired action. It is a well proven and popular concept in marketing and behavioral economics. The inventor of nudge - Dr Richard Thaler won Nobel prize in behavioral economics for his work on nudges. - Done my research on what key brands like Apple, Target, BestBuy, Amazon and others do, I've gone through external and internal research on nudges before coming to this forum. - Nudges are meant to boost user confidence, essentially telling the users, 'hey, it's alright, you're doing just fine, here's a piece of information to show it, and you can feel confident to proceed'. - Why we care about nudges is because our customers are complaining that they need more help in finding relevant products, configuring and comparing them. - If you think about it, they're looking for relevant nudges that can guide them further. Now, nudge marketing is an established tactic to boost user confidence to take desired action. All key brands, including premium brands like Apple, use nudges to direct user behavior. In Dell, it has mostly been static text used to manually create, for example badges, offer banners, etc. It has shown positive signs in recent experiments. - With the help of transact data and user context, it is possible to identify intent, then use appropriate nudge using near-real time insights. Let me show you how it can look like and then we can talk about why we are suggesting this and what potential we see based on available research. ##### Demo slides with animations - Walk them through slide animations. Notice how the nudge changes based on context, also notice how it gives relevant information in a non-stressful way. ##### Exec summary slide - There are 3 key datapoints out there - 2 from Dell and one from external research that all point to customers asking for help with product insights, searching the right product and figuring out the right configuration and then using dynamic nudges to help boost their confidence to convert. - Our internal Dell experiment on badges was done recently and it showed promise with recommendation from analytics team to beef up badge information with insights to improve conversions. - In fact, we see 20%-30% lift in external research, & conservatively at 4% lift, it'll give incremental XX revenue. We see most of our major ecommerce players like Amazon, Walmart and competitors like Apple, doing this. - To achieve these goals, what we are proposing here is following system: transact insights that takes user behavior context, product and inventory data to predict which nudge is most relevant, then give it to UX to display it in the optimum design and use a feedback loop from conversions of those users to move down the purchase funnel. - Such a system will get smarter over time as it gets this feedback so you can imagine that we may start with our best hypothesis on given set of nudges for a page and then let system use conversion data to start predicting. - What we are hoping from today’s discussion is to validate this proposal and get funding so that it becomes a serious outcome for us all. - I have several backup slides that I’m happy to walk through to address any questions. Take a pause. What would Dell mean for nudges. Next best action What is Dell's brand perception in customer's head Second - from customer's standpoint, what would make user go forward. Third - Tactically, what kind of nudges make sense from business standpoint We didn't have many doorbusters this time, mainly because Dell is positioning similar to Apple. Good homework on nudge vs FOMO Connect it with clean vision of dell.com Introduce it as a value prop, down to product visual - make that narrative I've already done homework of what other companies are doing. ### Feedbacks ##### Bree - - What's the sizing? - dell.com cart UX - Unified Cart Sizing - 4 sprints, 2 dev - Assuming new API integration, new API contract - Transact insights system - Harsh - This is how I did the research, how else? What do you think I should do different? What are my blind spots? I want to do more such case studies to present in front of my stakeholders like DTX and your org. ##### Harsh, Sarika - Crawl phase High promotion, high inventory Start with deals, Social proof Augment with persona flavor Saurabh work with Efrat, Aditya to work on customer problems - they don't know what's right for them. Can we sign up for numbers love the solutioning taylor it to the customer problems want all PLLs to think in this direction ##### Sai - Add flavor of persona to make sure we solve problems relevant to customers. ##### Justin Walker, - Claim bar has been relevant, scarcity tactics work - Shop path badging was experimented earlier - More impactful when they're limited to certain products. - Money, bundling, new, top rated ##### Jeremy - TnC testing needs to happen first. - Deals FOMO features - can start with this - Crawl is to experiment with what we have - Get metric for Deals from Jeremy - Jeremy to set up call with TnC lead - Ask Susana to see if there's any other test - tool to search through old tests ##### Prateek Srivastava - Change narrative around nudges – - Identify user problems  - Enable users do what they want to do. - Start with your space ### Research material [[Effect of digital nudges.pdf]]